Human Eye Model!

In the previous post I gave you all a glimpse of business application of Artificial Intelligence.
Computer vision was one such topic that i raised!

One of the advancement in computer vision with deep learning is called as “Convolutional neural network”.

There is a quote in the movie –“The man who knew infinity”
There is a dialogue in the movie:
"We are merely explorers of infinity in the pursuit of absolute perfection. We do not invent formulaes, they already exist in the nature we are only exploring them"



Convolutional neural network is a replica of Human Eye Model .
Convolutional neural network (CNN) is all about Image Processing and Analytics.
The agenda of CNN is to make the machines view the world as human beings.

Today we will not explore CNN rather since CNN Model is inspired by the human eye, we will explore the various ways in which a human eye process an Image.

Universe is made up of Electromagnetic Radiation (Energy + Wavelength + Frequency)
Frequency is measured in cycles per second, or Hertz. Wavelength is measured in meters. Energy is measured in electron volts.
The image below shows where you might encounter each portion of the EM spectrum in your day-to-day life.







As seen above, the visible spectrum of Electromagnetic spectrum is called as – “LIGHT” which we will study in detail today.
NASA definition of LIGHT: LIGHT is the common term for electromagnetic radiation, usually referring to that portion visible to the human eye. However other bands of the EM spectrum are also often referred to as different forms of light.

The main source of light on earth is SUN. Other sources of light includes Lamps, fire, power systems, electric lightsSome species like fireflies and vampire squids generate their own light.

A light consist of spectrum of all colours.





Light is made up of wave lengths ranging from Violet 400 nm to Dark Red 700 nm, each wave length corresponds to a particular colour as seen below





CNN works exactly the way a Human Eye works which means - your eyes are running CNN Models every time you observe or see something?
Today we are applying this Human Eye Model to computer vision.


Let us study Human Eye Model in detail

Our understanding of light and colour began with Isaac Newton (1642-1726)
Newton observed that colour is not inherent in objects, the surface of the object:
1) Reflect some colours – Visible to Human Eye
2) Absorbs some colours 

Thus, red is not "in" an apple. The surface of the apple is reflecting the wavelengths we see as red and absorbing all the rest of the wavelength.
An object appears white when it reflects all wavelengths and black when it absorbs them all.




Red, green and blue (RGB) are the primary colours of the colour spectrum ,by varying the amount of red, green and blue light, all of the colours in the visible spectrum can be produced by the human eye.




Human eye and how do we see light?

Do you know a human eye can perceive about 382000 different colours?
How do we do it?

Human eye is lined with a thin layer called the retina.
Photo-receptors are located in the retina
There are 2 types of photo receptors –Rods and Cones
If eye was a camera, the retina would be the film.





Rods: GREY
Rods are responsible for vision at low light levels.
We use Rods for night vision, Rods don’t help In colour vision, in the night we see everything in GREY SCALE
Human eyes has 100 million rod cells

Cones: RED, BLUE, GREEN
There are 3 types of Cone: RED, BLUE, GREEN
Cones help us visualise the detail of any image.
Variety of colours are produced with by mixing RED, BLUE, and GREEN
Human eyes has 6 million Cones 






The human eye and brain together translate light into colour.  The colour perceived by human is the sum of:
          the energy spectrum of the light
          the reflection spectrum of the object
          the response spectrum of the eye

382000 different colours are perceived by human eye by some kind of colour mixing classification system. Colour mixing Model of the Human Eye can be classified into following types.
There has been some recent innovations about which iam not writing now.



1)      CIE(commission Internationale de l’Ḗclairage) standard
Ø   Cannot be replicated by a computer model 
Ø   each colour = a weighted sum of three imaginary primary colours, meaning you can simply imagine 3 colours and mix them to produce a new colour inside your eye. This is called as ”Power of Imagination” which is something that CNN is not capable of doing and only human eye can do.

2)      RGB colour model - RED , BLUE ,GREEN.
Ø  all colours perceived by the human eye are generated from the above three primaries
Ø  various colours are obtained by changing the amount of each primary
Ø  additive mixing (r,g,b), 0≤r,g,b≤1


     


3)      CMY colour model (also, CMYK)
Ø  cyan, magenta and yellow are complimentary colours of red,green and blue, respectively
Ø  subtractive mixing
Ø  For printing and graphics art industry, CMY is not enough; a fourth primary, K which stands for black, is added.


HSL (hue, saturation, lightness) and HSV (hue, saturation, value) are alternative representations of the RGB colour model, designed in the 1970s by computer graphics researchers to more closely align with the way human vision perceives colour-making attributes.


4)      HSV colour model separates Image intensity from colour formation which is absent in RGB Model. Hence many Computer Vision Algorithms  uses this Model for image processing and detection.(Better fit for Clinical Trial Imaging)
Ø  HSV stands for Hue-Saturation-Value
Ø  described by a hexcone derived from the RGB cube


5)      HSL colour model
Ø  Another model similar to HSV is HSL
Ø  HSL stands for Hue-Saturation-Lightness


(There is also GREY SCALE MIXING = Rods) 

Every Artificial Intelligent Algorithm comes from Nature's Intelligence.
In the coming days we will explore the origin of each of the Artificial Intelligence Algorithms and how and why they are applied.
As iam not an engineer i will not be able to provide information about coding , but yes economic aspect of these will be discussed in great detail.

In my next post i will explain CNN in the form of conceptual design required for business application.

Comments